pydantic_settings: parse list from the environment variables

[stdedos]

I have the following in .env file:

export A_IDS=1228

and

    a: list[int]

but I am getting

a
  Input should be a valid list [type=list_type, input_value=1228, input_type=int]
    For further information visit https://errors.pydantic.dev/2.12/v/list_type

I looked at

• https://docs.pydantic.dev/latest/concepts/pydantic_settings/#parsing-environment-variable-values
• https://github.com/pydantic/pydantic/discussions?discussions_q=pydantic_settings+list

but I didn't find anything related to my query.

There is even a related example:

os.environ['numbers'] = '1,2,3'
print(Settings().model_dump())
#> {'numbers': [1, 2, 3]}

my only difference is: for now, I have only one value.

I tried to cheat with 1228, (aka the Python way of enforcing a tuple), but that explodes even worsely.

json.decoder.JSONDecodeError: Extra data: line 1 column 5 (char 4)

[janas-adam]

hi @stdedos

u can try:

# .env
A=asdasd
ASD="[123]"

# settings
class Settings(BaseSettings):
    a: str
    asd: list[int]

    model_config = SettingsConfigDict(
        env_file=".env",      
        case_sensitive=False
    )

result:

>> Settings().model_dump()
`{'asd': [123], 'a': 'asdasd'}`

hope it ll solve ur problem

[ezsouza]

Hey @stdedos! The doc example uses 1,2,3 because pydantic_settings splits by comma by default. When you have only one value like A_IDS=1228, you need to wrap it in brackets so it parses as a list.

The cleanest way is to change your .env to:

export A_IDS="[1228]"

This makes pydantic_settings treat it as a JSON array with one element. The response above with "[123]" is on the right track.

But if you really need to keep the format as A_IDS=1228 without brackets, you can use a field validator:

from pydantic import BaseModel, field_validator
from pydantic_settings import BaseSettings

class Settings(BaseSettings):
    a_ids: list[int]
    
    @field_validator('a_ids', mode='before')
    @classmethod
    def parse_single_to_list(cls, v):
        if isinstance(v, str):
            # Check if it's already a bracketed list like "[123]"
            if v.startswith('['):
                return v
            # Otherwise wrap single values in brackets
            return f"[{v}]"
        return v

settings = Settings()

That way A_IDS=1228 gets auto-wrapped to [1228] before pydantic parses it.

Let me know if it works!

[stdedos]

Indeed it works - thank you @ezsouza (also @janas-adam was technically first 😊)

However both of them work only for numbers, not strings (which I assume they need to be quoted) - and neither does export A_IDS= work with it either.

I guess going with the enable_decoding=False, and hand-rolled parsing would be the cleanest way (to avoid embedding unnecessary Json)